Efficient Forecasting of Large-Scale Hierarchical Time Series via Multilevel Clustering †
Abstract
:1. Introduction
2. Backgrounds
3. Hierarchical Time Series Clustering
3.1. Two-Level Time Series Clustering
3.2. Multilevel Time Series Clustering
Algorithm 1 HTS-Cluster. |
|
4. Experiments
4.1. HTS Clustering
4.2. HTS Forecasting
5. Conclusions
Author Contributions
Funding
Data Availability Statement
Conflicts of Interest
References
- Wickramasuriya, S.L.; Athanasopoulos, G.; Hyndman, R.J. Optimal forecast reconciliation for hierarchical and grouped time series through trace minimization. J. Am. Stat. Assoc. 2019, 114, 804–819. [Google Scholar] [CrossRef]
- Petitjean, F.; Ketterlin, A.; Gançarski, P. A global averaging method for dynamic time warping, with applications to clustering. Pattern Recognit. 2011, 44, 678–693. [Google Scholar] [CrossRef]
- Aghabozorgi, S.; Shirkhorshidi, A.S.; Wah, T.Y. Time-series clustering–a decade review. Inf. Syst. 2015, 53, 16–38. [Google Scholar] [CrossRef]
- Zhong, S.; Ghosh, J. A unified framework for model-based clustering. J. Mach. Learn. Res. 2003, 4, 1001–1037. [Google Scholar]
- Ma, Q.; Zheng, J.; Li, S.; Cottrell, G.W. Learning representations for time series clustering. Adv. Neural Inf. Process. Syst. 2019, 32, 3781–3791. [Google Scholar] [CrossRef]
- Ho, N.; Nguyen, X.; Yurochkin, M.; Bui, H.H.; Huynh, V.; Phung, D. Multilevel clustering via Wasserstein means. In Proceedings of the International Conference on Machine Learning, PMLR, Sydney, Australia, 6–11 August 2017; pp. 1501–1509. [Google Scholar]
- Ho, N.; Huynh, V.; Phung, D.; Jordan, M. Probabilistic multilevel clustering via composite transportation distance. In Proceedings of the 22nd International Conference on Artificial Intelligence and Statistics, PMLR, Okinawa, Japan, 16–18 April 2019; pp. 3149–3157. [Google Scholar]
- Rodrigues, P.P.; Gama, J.; Pedroso, J. Hierarchical clustering of time-series data streams. IEEE Trans. Knowl. Data Eng. 2008, 20, 615–627. [Google Scholar] [CrossRef]
- Blondel, M.; Mensch, A.; Vert, J.P. Differentiable divergences between time series. In Proceedings of the International Conference on Artificial Intelligence and Statistics, PMLR, Virtual, 13–15 April 2021; pp. 3853–3861. [Google Scholar]
- Müller, M. Dynamic time warping. Inf. Retr. Music. Motion 2007, 69–84. [Google Scholar]
- Cuturi, M.; Blondel, M. Soft-dtw: A differentiable loss function for time-series. In Proceedings of the International Conference on Machine Learning, PMLR, Sydney, Australia, 6–11 August 2017; pp. 894–903. [Google Scholar]
- Cuturi, M.; Doucet, A. Fast computation of Wasserstein barycenters. In Proceedings of the International Conference on Machine Learning, PMLR, Beijing, China, 21—26 June 2014; pp. 685–693. [Google Scholar]
- Schütze, H.; Manning, C.D.; Raghavan, P. Introduction to Information Retrieval; Cambridge University Press: Cambridge, UK, 2008; Volume 39. [Google Scholar]
- Hubert, L.; Arabie, P. Comparing partitions. J. Classif. 1985, 2, 193–218. [Google Scholar] [CrossRef]
- Vinh, N.X.; Epps, J.; Bailey, J. Information theoretic measures for clusterings comparison: Variants, properties, normalization and correction for chance. J. Mach. Learn. Res. 2010, 11, 2837–2854. [Google Scholar]
- Hyndman, R.J.; Koehler, A.B. Another look at measures of forecast accuracy. Int. J. Forecast. 2006, 22, 679–688. [Google Scholar] [CrossRef] [Green Version]
- Salinas, D.; Flunkert, V.; Gasthaus, J.; Januschowski, T. DeepAR: Probabilistic forecasting with autoregressive recurrent networks. Int. J. Forecast. 2020, 36, 1181–1191. [Google Scholar] [CrossRef]
- Lai, G.; Chang, W.C.; Yang, Y.; Liu, H. Modeling long-and short-term temporal patterns with deep neural networks. In Proceedings of the 41st International ACM SIGIR Conference on Research & Development in Information Retrieval, Ann Arbor, MI, USA, 8–12 July 2018; pp. 95–104. [Google Scholar]
- Makridakis, S.; Spiliotis, E.; Assimakopoulos, V. The M5 accuracy competition: Results, findings and conclusions. Int. J. Forecast. 2022, 38, 1346–1364. [Google Scholar] [CrossRef]
Method∖Metric | Time (s) | Global | Local | ||||
---|---|---|---|---|---|---|---|
NMI | AMI | ARI | NMI | AMI | ARI | ||
DTCR | 132 | 0.325 ± 0.012 | 0.257 ± 0.023 | 0.21 ± 0.011 | 0.392 ± 0.014 | 0.313 ± 0.006 | 0.284 ± 0.009 |
Soft-DTW | 67 | 0.412 ± 0.009 | 0.326 ± 0.019 | 0.277 ± 0.008 | 0.411 ± 0.022 | 0.342 ± 0.009 | 0.304 ± 0.014 |
Concat | 186 | 0.436 ± 0.015 | 0.342 ± 0.014 | 0.314 ± 0.016 | 0.411 ± 0.022 | 0.342 ± 0.009 | 0.304 ± 0.014 |
HTS-Cluster | 37 | 0.455 ± 0.018 | 0.354 ± 0.015 | 0.302 ± 0.013 | 0.424 ± 0.018 | 0.366 ± 0.013 | 0.321 ± 0.018 |
DTCR | 72 | 0.065 ± 0.002 | 0.015 ± 0.001 | 0.008 ± 0.002 | 0.105 ± 0.011 | 0.059 ± 0.002 | 0.054 ± 0.003 |
Soft-DTW | 49 | 0.119 ± 0.005 | 0.043 ± 0.003 | 0.027 ± 0.003 | 0.126 ± 0.008 | 0.082 ± 0.006 | 0.061 ± 0.005 |
Concat | 174 | 0.135 ± 0.004 | 0.073 ± 0.007 | 0.045 ± 0.006 | 0.126 ± 0.008 | 0.082 ± 0.006 | 0.061 ± 0.005 |
HTS-Cluster | 34 | 0.134 ± 0.005 | 0.075 ± 0.005 | 0.041 ± 0.004 | 0.128 ± 0.014 | 0.064 ± 0.005 | 0.065 ± 0.002 |
Level | Metric | Simulation | Financial Record | ||||||
---|---|---|---|---|---|---|---|---|---|
DTCR | Soft-DTW | Concat | HTS-Cluster | DTCR | Soft-DTW | Concat | HTS-Cluster | ||
1 | NMI | 0.28 | 0.313 | 0.342 | 0.356 | 0.037 | 0.124 | 0.156 | 0.154 |
AMI | 0.243 | 0.277 | 0.301 | 0.322 | 0.021 | 0.079 | 0.112 | 0.106 | |
ARI | 0.221 | 0.265 | 0.285 | 0.304 | 0.009 | 0.056 | 0.094 | 0.092 | |
2 | NMI | 0.298 | 0.317 | 0.357 | 0.375 | 0.056 | 0.116 | 0.147 | 0.152 |
AMI | 0.271 | 0.282 | 0.314 | 0.346 | 0.034 | 0.087 | 0.115 | 0.121 | |
ARI | 0.236 | 0.259 | 0.302 | 0.317 | 0.016 | 0.034 | 0.083 | 0.092 | |
3 | NMI | 0.272 | 0.324 | 0.364 | 0.372 | 0.055 | 0.134 | 0.163 | 0.172 |
AMI | 0.234 | 0.295 | 0.322 | 0.33 | 0.028 | 0.098 | 0.132 | 0.141 | |
ARI | 0.217 | 0.268 | 0.307 | 0.309 | 0.012 | 0.057 | 0.106 | 0.113 | |
4 | NMI | 0.303 | 0.369 | 0.369 | 0.369 | 0.076 | 0.136 | 0.136 | 0.136 |
AMI | 0.275 | 0.341 | 0.341 | 0.341 | 0.043 | 0.102 | 0.102 | 0.102 | |
ARI | 0.264 | 0.316 | 0.316 | 0.316 | 0.026 | 0.061 | 0.061 | 0.061 |
Method/Level | 1 | 2 | 3 | 4 | Total Time | |
---|---|---|---|---|---|---|
Without cluster | 62.39 | 76.26 | 78.25 | 84.14 | 1 | |
DTCR | 82.35 | 96.09 | 104.85 | 104.33 | 0.39 | |
Soft-DTW | 78.61 | 93.04 | 93.12 | 96.76 | 0.27 | |
Concat | 74.24 | 84.65 | 83.73 | 96.76 | 0.57 | |
HTS-Cluster | 72.99 | 80.07 | 85.29 | 96.76 | 0.16 |
Dataset | Wiki | M5 | ||||||||
---|---|---|---|---|---|---|---|---|---|---|
Levels | 1 | 2 | 3 | 4 | 5 | 8 | 9 | 10 | 11 | 12 |
LSTNet | 76.36 | 76.89 | 79.65 | 81.13 | 86.22 | 63.74 | 69.43 | 73.35 | 76.46 | 82.36 |
LSTNet-Cluster | 76.33 | 76.56 | 77.68 | 78.07 | 95.16 | 62.48 | 69.14 | 71.11 | 76.52 | 98.78 |
DeepAR | 73.98 | 74.54 | 77.42 | 79.12 | 84.77 | 59.36 | 67.18 | 72.04 | 76.41 | 80.24 |
DeepAR-Cluster | 74.21 | 74.37 | 77.36 | 77.56 | 89.67 | 58.74 | 65.46 | 74.39 | 75.04 | 90.49 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Han, X.; Ren, T.; Hu, J.; Ghosh, J.; Ho, N. Efficient Forecasting of Large-Scale Hierarchical Time Series via Multilevel Clustering. Eng. Proc. 2023, 39, 31. https://doi.org/10.3390/engproc2023039031
Han X, Ren T, Hu J, Ghosh J, Ho N. Efficient Forecasting of Large-Scale Hierarchical Time Series via Multilevel Clustering. Engineering Proceedings. 2023; 39(1):31. https://doi.org/10.3390/engproc2023039031
Chicago/Turabian StyleHan, Xing, Tongzheng Ren, Jing Hu, Joydeep Ghosh, and Nhat Ho. 2023. "Efficient Forecasting of Large-Scale Hierarchical Time Series via Multilevel Clustering" Engineering Proceedings 39, no. 1: 31. https://doi.org/10.3390/engproc2023039031